SDK auto-instrumentation ergonomics: Routing AI calls to specific projects #1522
Description
This is a bit of an open question: How should you be able to route individual LLM calls / agents to various different braintrust projects?
Current ideas:
- Provide a primitive that takes a callback wherein all traced calls are routed to a specific project or with a specific configuration. e.g.
await withBraintrustProject('project-a', async () => { ... })await braintrust.withConfiguration({ project: 'project-b', metadata: { foo: 'bar' } }, async () => { ... })